Our AI Future: AI Is Disrupting The Insurance Industry But New, Ethical, Business Models Are Possible
The insurance industry doesn't have a good reputation because the way it makes money is by spending as little as possible on claims, which causes friction with the clients it is supposed to be helping. Now AI systems are also in the mix, which could be a disaster if they are implemented unethically, but one company has decided to use the technologies to the benefit of both itself and the customers it serves.
This is the fourth in a series of articles highlighting interesting technologies and concepts that caught my attention at the inaugural AI Expo Africa, which took place in Cape Town last year. The first article covers disruption
(and volumetric holograms), the second looks at customer experience
(and chatbots), the third explores synergy
(and automation), and this one looks at new business models (and the insurance industry).
Insurance is a very unpleasant business. As a client you feel that you have to pay a fortune in premiums and then, when it comes time to claim, it's often a tedious fight as the insurance company obfuscates the "rules" and tries to strongarm you into a deal that'll see it paying out as little as possible. (Can you tell I'm talking from experience?) Add AI to the mix in ways that are opaque to the average consumer and it's a terrifying prospect. In many cases it's actually just used to reduce costs, speed up processes, and help with client control but where it gets tricky is what happens beyond that as it gets integrated into more and more processes and systems.
"AI is fundamentally going to change the insurance industry, and it will never be the same again," Alex Thomson, the co-founder of Naked Insurance, told delegates at the inaugural AI Expo Africa in Cape Town last year. This is for two reasons: the existing players that aren't in a good position to pivot along with these changes are going to be left behind and those who are able to integrate artificial intelligence systems smartly will face an ethical choice as to whether they do it malevolently or benevolently.
"What seemed to us to be fairly clear was that the businesses that would win in the future would use AI and modern tech not as a kind of bolt on onto a legacy business model but would design from scratch with AI and modern tech in mind, and that's what gave rise to us starting Naked," Thomson said. This realisation, about four years ago, resulted in the company kicking off towards the end of 2016 and then launching its first product - car insurance - with a backend that is built from scratch.
Before we look at how Naked Insurance approaches insurance we need to understand how the industry works now because, as Thomson said, "insurance as a product or service is pretty unusual". This is because usually when you buy a product you can first test it out - you can take a car for a test drive, you can sit on a chair, you can read reviews to see what other people's experiences have been and how it might relate to your experience. It's a relatively effective way to assess the quality of something you have no experience with and manufacturers and suppliers therefore have an incentive to invest in quality.
This is in sharp contrast to how insurance works.
"In insurance there's very, very little objective evidence that you can actually use to decide which insurance product to buy. Your ability to evaluate that product is highly compromised. You can wait for ages, you pay and you pay and you pay and you pay, you have no idea if that thing is actually going to deliver what you thought it was until you claim and that could be months or years down the line. [...] You can't purchase insurance from a really informed position. [...] The performance of the product can only be assessed in the future. [...] The performance of the product depends on the provider's (insurance company's) discretion because there are grey areas in insurance wording. The rules change and each claim handler handles it differently. It's very, very difficult to get a clear picture of that. The problem with this discretion is that there is also a conflict of interest in play there. When an insurer pays a claim it comes out of its profits; it makes less money. Conversely, if it can find a way not to pay a claim, it makes more money.
"Insurers really have free reign. If they do bad things there's very little control of it. There's no market force anywhere that is going to bring that back and shut down that behaviour. It's that that is really at the core of the problem of insurance - and I'm going to link that into AI. The two key things that we see going wrong is this conflict behaviour at claims stage [...] and then overcharging on renewals.
"So what about AI in this world? Is AI going to address all of these things - are we suddenly going to have an industry that's behaving very well? If what I've said has made any sense it's actually probably geared up the other way. AI can be used for all sorts of nefarious means to try to grow shareholder profit really at the expense of customer outcomes meaning that insurance can stray further from its sort of social purpose than it is even now. Take for example the conflict at claims stage. People [such as call-centre agents] are pretty good at evaluating 'how can I make you go away?' [when dealing with claimants]. An AI can get even better at that, predicting, you know, are you the kind of person that if I strongarm you a bit you'll run off - 'I think you were driving really badly when you had that kind of accident - I'm not sure we're going to pay that'. An AI can get incredibly good at classifying those people, incredibly good at figuring out the right mechanism for making people go away, for making them feel like it's their fault. I think there's significant risk of AI being used for that.
"So when we looked at all of this we felt a more fundamental change was needed to be able to clean up the industry and to be able to support AI that works for the benefit of customers and the benefit of the functioning of insurance and hopefully to also diminish the amount of regulation that we see in the industry. So what we've done is change the way we make money.
In the traditional business model for insurance a company makes money based on the underwriting profits so it will do better if the claims are low and worse if the claims are high. There is, therefore, an incentive to do anything to reduce the claims. Naked Insurance, on the other hand, uses a different business model: it charges a flat 20% fee off the premium and that is the total income that the company generates. At the end of the year, if there is any money left over, it is donated to charities of customers' choosing.
"What [this business model] means is that when we're building AI we're not trying to grow underwriting profits, we're trying to look after the relationship with our clients because the longer they stay around, the better we do," Thomson said. "That's our primary
objective, rather than trying to grow our underwriting margin. We still have a responsibility to the underwriting pool to make sure that we're charging fairly, the valid claim's getting paid, and so on, but we don't have to think about, you know, 'how can I grow our underwriting margin from 6% to 7% this year by doing some suspect things?'."
The Naked Insurance experience is very different for customers too, from the moment potential clients interact with the company for the first time for a quote, to the claims process. Getting a quote is a quick Q&A session with a chatbot
via the web site or the app, which is available for both iOS
. Once on board customers then submit supplemental data about a car - videos and photographs - via the app.
"Probably the best process, prior to us launching, was really about a 40-minute phone call to get a final quote for car insurance," Thomson said. Instead the company is "just making it quick, and easy, and pretty simple, and really just taking the control away from the call centre and putting it into the hands of the customer".
Above: To add a car to a Naked Insurance policy users interact with the system via a chatbot and take photographs and video of their car, which they submit via the app.
One of the ways in which the company saves time and reduces costs is to minimise the number of communications options it offers - customers can only sign up on the web site or via the app. It's a much narrower set of options than what other insurance companies offer, which have legacy methods in place, including face to face and via a call centre, that additionally result in more varied, less streamlined data being gathered.
"What we have here is one example of a consistent approach, which builds a consistent data set, that allows us to build models more easily," Thomson said. Customers can also manage almost everything via the app, although there are humans available via email for more complicated tasks or urgent assistance.
The claims process is also very different. Once again customers use the app to document and submit data, which will include a video and photographs, and, depending on the complexity of the problem, the claim can be approved immediately. Not all claims are handled this way yet - it's usually simple claims that will experience this quick approval - but the intention in the long term is for the majority of claims to be handled this way, which will come about as the company's data grows and its AI systems become more sophisticated.
One of the reasons that more complicated claims are not yet approved instantly is due to the industry-wide problem of fraud at the claims stage but it's another area that will increasingly be tackled with AI systems - and Naked Insurance is no exception. "What we're trying to do with AI is be able to use new sources of data and some really interesting analytics to pick up when there are signs of dishonesty and to pick up when a claim might not be valid," Thomson said before pointing out that one of the areas that the company is researching is indicators of lying that can be picked up in the submission videos.
Fraud is not the only area that is benefitting from increasingly sophisticated systems, however.
"When we look at insurance pretty much every step of the chain has opportunities to be transformed radically with AI," Thomson said. "One of the areas that's moved really quickly over the last year or two both locally and internationally has been been in the area of automated repair estimation. Tools are available right now [in which] feeding in a bunch of pictures will outperform a human assessor and will be able to do the classification of the images of the different parts of the car, determine the damage, and actually estimate the cost of the repair in terms of what parts and labour [are needed]."
From a technology standpoint the future of the insurance industry is very exciting but for the average customer the future is still one of great uncertainty. I'll let Thomson have the last word on that: "We see a very positive future for the industry if [AI integration] is dealt with in the right way and an industry where we can see that use of automation to lower costs and put the control into the customer's hands, where we have a lot more objectivity and less subjectivity due to the use of AI and ultimately a better relationship with our clients and a product where people feel that it's going to deliver what they expect of our system," he said.
Let's hope that the Naked Insurance ethos, which seems to be an outlier in the insurance space, ultimately becomes the industry's operating norm.
Mandy J Watson was a media guest at the 2018 AI Expo Africa conference. The 2019 conference is being held in Cape Town at the Century City Conference Centre on 4 and 5 September 2019.